Personalized Content Generator

A web application that generates content — blog posts, emails, social media posts, essays, and more — that sounds exactly like you wrote it. It creates a "digital twin" of your writing style by analyzing your existing writing samples using RAG (Retrieval-Augmented Generation) and style profiling via the Claude API.

No custom model training required. It uses your writing samples as few-shot context alongside a structured style profile to produce content indistinguishable from your own voice.

How It Works

1. Upload your writing samples — past blog posts, emails, social media posts, essays, etc.
2. Generate your style profile — Claude analyzes your writing to extract tone, vocabulary, sentence structure, quirks, and rhetorical patterns
3. Generate content — Describe what you want to write; the app retrieves your most relevant writing samples and uses them as examples to generate content in your exact voice

Features

• Upload files (.txt, .md, .pdf, .docx) or paste text directly
• Automatic style profile extraction (tone, formality, vocabulary, sentence patterns, quirks)
• RAG-powered generation: retrieves your writing samples relevant to the topic before generating
• Regenerate or refine any piece of content
• Rate generated content (1–5 stars)
• Full generation history
• Clean web UI — no Node.js required

Tech Stack

Layer	Technology
Backend	Python 3.11+ / FastAPI
Database	SQLite (WAL mode) / SQLAlchemy async
Vector DB	ChromaDB (embedded, local embeddings)
LLM	Claude API (Sonnet for generation, Haiku for analysis)
Frontend	Jinja2 + HTMX + Tailwind CSS (CDN)
File parsing	pymupdf (PDF), python-docx (DOCX)

Setup

1. Clone and create a virtual environment

git clone <repo-url>
cd "Personalized Content Generator"
python3 -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

2. Install dependencies

pip install -e ".[dev]"

3. Configure environment

cp .env.example .env

Edit .env and add your Anthropic API key:

ANTHROPIC_API_KEY=sk-ant-your-key-here

Get a key at console.anthropic.com.

4. Run the server

uvicorn app.main:app --reload

Open http://127.0.0.1:8000 in your browser.

Usage

Step 1 — Add writing samples

Go to Writing Samples and either:

• Upload .txt, .md, .pdf, or .docx files (max 10MB each)
• Paste text directly with a title

Add as many samples as you have — more samples = better style matching. Label each with the content type (blog, email, social, etc.).

Step 2 — Generate your style profile

Go to Dashboard and click Generate Profile. Claude will analyze your samples and extract a detailed style profile. This takes ~10–30 seconds. You can regenerate it anytime as you add more samples.

Step 3 — Generate content

Go to Generate and fill in:

• Content type (blog, email, tweet, LinkedIn post, essay)
• Topic
• Target audience (optional)
• Length (short / medium / long)
• Any additional instructions

Click Generate. If the result isn't quite right, click Regenerate for a different version or use Refine to give specific feedback.

API

All functionality is available as a REST API:

Method	Endpoint	Description
POST	/api/samples/upload	Upload files
POST	/api/samples/paste	Submit pasted text
GET	/api/samples	List samples
DELETE	/api/samples/{id}	Delete a sample
GET	/api/profile	Get active style profile
POST	/api/profile/generate	Generate/regenerate profile
POST	/api/generate	Generate content
POST	/api/generate/{id}/regenerate	Regenerate
POST	/api/generate/{id}/refine	Refine with instructions
GET	/api/history	List generation history
POST	/api/history/{id}/rate	Rate (1–5)

Interactive API docs available at http://127.0.0.1:8000/docs.

Estimated API Costs

Operation	Model	Approx. Cost
Style analysis	Claude Haiku 4.5	~$0.01 per run
Generate (short)	Claude Sonnet 4	~$0.02
Generate (medium)	Claude Sonnet 4	~$0.03
Generate (long)	Claude Sonnet 4	~$0.05
Embeddings	Local (MiniLM)	Free

Light personal use (~50 generations/month): ~$1–3/month.

Privacy

• Writing samples are sent to the Anthropic API for style analysis and generation
• No data is stored by Anthropic beyond their standard API retention policy
• All your data (samples, profile, history) is stored locally in data/
• Your .env file (containing your API key) is gitignored

Project Structure

app/
├── main.py                  # FastAPI app, lifespan, startup reconciliation
├── config.py                # Settings via environment variables
├── api/                     # JSON REST API endpoints
├── pages/                   # HTML page routes
├── models/                  # SQLAlchemy ORM models + Pydantic schemas
├── services/
│   ├── file_parsers.py      # Safe file parsing with limits
│   ├── ingestion.py         # Text chunking for RAG
│   ├── vectorstore.py       # ChromaDB wrapper
│   ├── style_analyzer.py    # Style profile extraction
│   └── generator.py         # RAG + prompt assembly + generation
├── prompts/                 # Prompt templates for Claude
└── templates/               # Jinja2 HTML templates
data/                        # Runtime data (gitignored)
├── content_gen.db           # SQLite database
├── chroma/                  # ChromaDB vector store
└── uploads/                 # Temporary upload storage